clinical article
Radiosurgery for temporal lobe arteriovenous malformations: effect of temporal location on seizure outcomes Dale Ding, MD,1 Mark Quigg, MD, MSc,2 Robert M. Starke, MD, MSc,1 Zhiyuan Xu, MD,1 Chun-Po Yen, MD,1 Colin J. Przybylowski, BS,1 Blair K. Dodson, BSN,1 and Jason P. Sheehan, MD, PhD1 Departments of 1Neurological Surgery and 2Neurology, University of Virginia, Charlottesville, Virginia
Object The temporal lobe is particularly susceptible to epileptogenesis. However, the routine use of anticonvulsant therapy is not implemented in temporal lobe AVM patients without seizures at presentation. The goals of this case-control study were to determine the radiosurgical outcomes for temporal lobe AVMs and to define the effect of temporal lobe location on postradiosurgery AVM seizure outcomes. Methods From a database of approximately 1400 patients, the authors generated a case cohort from patients with temporal lobe AVMs with at least 2 years follow-up or obliteration. A control cohort with similar baseline AVM characteristics was generated, blinded to outcome, from patients with non-temporal, cortical AVMs. They evaluated the rates and predictors of seizure freedom or decreased seizure frequency in patients with seizures or de novo seizures in those without seizures. Results A total of 175 temporal lobe AVMs were identified based on the inclusion criteria. Seizure was the presenting symptom in 38% of patients. The median AVM volume was 3.3 cm3, and the Spetzler-Martin grade was III or higher in 39% of cases. The median radiosurgical prescription dose was 22 Gy. At a median clinical follow-up of 73 months, the rates of seizure control and de novo seizures were 62% and 2%, respectively. Prior embolization (p = 0.023) and lower radiosurgical dose (p = 0.027) were significant predictors of seizure control. Neither temporal lobe location (p = 0.187) nor obliteration (p = 0.522) affected seizure outcomes. The cumulative obliteration rate was 63%, which was significantly higher in patients without seizures at presentation (p = 0.046). The rates of symptomatic and permanent radiationinduced changes were 3% and 1%, respectively. The annual risk of postradiosurgery hemorrhage was 1.3%. Conclusions Radiosurgery is an effective treatment for temporal lobe AVMs. Furthermore, radiosurgery is protective against seizure progression in patients with temporal lobe AVM–associated seizures. Temporal lobe location does not affect radiosurgery-induced seizure control. The low risk of new-onset seizures in patients with temporal or extratemporal AVMs does not seem to warrant prophylactic use of anticonvulsants. http://thejns.org/doi/abs/10.3171/2014.10.JNS141807
T
KEY WORDS epilepsy; Gamma Knife; intracranial arteriovenous malformation; stereotactic radiosurgery; seizures; temporal lobe; vascular malformations; vascular disorders
of temporal lobe arteriovenous malformations (AVMs) carries burdens beyond the primary consideration of prevention of hemorrhage. Prior studies of AVM-associated epilepsy have shown that higher rates of epilepsy are associated with temporal lobe involvement.20,23,24,48 Interventional therapy has the potential to control AVM-associated epilepsy.2,23 Since postreatment
treatment epilepsy status has a strong influence on quality of life and functional outcomes, treatment-induced seizure control is an important consideration.25 Thus, the success of an intervention is dictated by more than obliteration of the AVM nidus. Radiosurgery is a minimally invasive treatment alternative to microsurgery for AVMs and induces gradual obliteration of the nidus over
ABBREVIATIONS AED = antiepileptic drug; AVM = arteriovenous malformation; MTLE = mesial temporal lobe epilepsy; RBAS = radiosurgery-based AVM score; RIC = radiation-induced change; VRAS = Virginia Radiosurgery AVM Scale. submitted August 6, 2014. accepted October 20, 2014. include when citing Published online April 17, 2015; DOI: 10.3171/2014.10.JNS141807. DISCLOSURE The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper. ©AANS, 2015
J Neurosurg April 17, 2015
1
D. Ding et al.
a typical period of 2 to 3 years.37,47 While AVM radiosurgery outcomes have been previously reported by our institution and other neurosurgical centers, the relationship between temporal lobe location and AVM radiosurgery outcomes, specifically with regard to seizure control, has not been previously defined.9–11,27,33 Furthermore, despite the particular susceptibility of patients harboring temporal lobe lesions to seizures, antiepileptic drug (AED) therapy is not routinely administered to patients with temporal lobe AVMs who do not have seizures at presentation.14 Therefore, the goals of the present study are 1) to evaluate the radiosurgery outcomes, with respect to seizure control, obliteration, and complications, for temporal lobe AVMs and 2) to determine the effect of temporal lobe location on seizure control following AVM radiosurgery.
Methods
Patient Selection This is a single-center, retrospective, case-control cohort study evaluated from a prospectively collected database, approved by the University of Virginia institutional review board, of over 1400 AVM patients treated with Gamma Knife radiosurgery from 1989 to 2012. The inclusion criteria for the case cohort were 1) temporal lobe AVM location, 2) data regarding presence or absence of seizures at presentation, 3) data regarding posttreatment seizure outcomes available, 4) complete AVM obliteration following radiosurgery, and 5) minimum posttreatment follow-up duration of 2 years for AVMs without obliteration. Patients who had more than 1 radiosurgery treatment were not excluded, but the duration of follow-up was calculated from the last procedure. Radiosurgery Procedure The Gamma Knife (Elekta) radiosurgery procedure was performed in a standard fashion as previously described.47 Prior to 1991, MRI was not routinely used in addition to angiography for treatment planning. After 1991, a combination of MRI and angiography were routinely used to optimize the spatial accuracy of radiosurgical planning. From 1989 until June 1994, the Kula software was used for dose planning. From July 1994 onward, dose planning was performed with the GammaPlan software (Elekta). Data and Variables The following variables were obtained from the database and directed chart review: patient demographic characteristics (sex, age, and presenting symptom), baseline AVM characteristics (prior hemorrhage, prior embolization, prior resection, maximum diameter, volume, eloquent location, location and number of draining veins, and presence of associated aneurysms), and radiosurgery characteristics (prescription dose, maximum dose, isodose line, and number of isocenters). The Spetzler-Martin grade, modified radiosurgery-based AVM score (RBAS), and Virginia Radiosurgery AVM Scale (VRAS) were determined for each nidus.45,46,49 To determine the effect of temporal lobe location on AVM-associated seizure outcomes following radiosurgery, a control cohort of patients with non-temporal corti2
J Neurosurg April 17, 2015
cal AVMs was generated from the same institutional database. The inclusion criteria for the control cohort were the same as those for the case cohort with the exception of location (i.e., frontal, parietal, or occipital lobe location for the control cohort vs temporal lobe location for the case cohort). To prevent statistical bias due to differences in baseline AVM characteristics and treatment parameters, the control cohort was selected so that the frequencies of preradiosurgery hemorrhage, preradiosurgery embolization and resection, eloquent location, presence of deep venous drainage, number of draining veins, and presence of associated aneurysms and the means of the nidus volume and prescription dose did not differ significantly between the temporal lobe AVM and non-temporal, cortical AVM cohorts. The generation of the control cohort was performed blinded to outcome and was conducted separately for AVM patients presenting with seizures and AVM patients not presenting with seizures. Radiological and Clinical Follow-Up Information regarding seizure status was extracted from the chart and categorized as seizure free (with seizure freedom being patient-defined absence of seizures) or as not seizure free (patient-defined improved, unchanged, or worsened seizure rates). Seizure control was defined, in patients with seizures at presentation, as seizure freedom or decreased seizure frequency regardless of AED therapy. Seizure worsening was defined as increased frequency or intensity of seizures regardless of AED therapy. For the purposes of data analysis, postradiosurgery seizure outcomes were dichotomized into those patients with seizure control and those patients without seizure control (i.e., unchanged seizure status or worsening). De novo seizures were defined, in patients without seizure presentation, as new-onset seizures occurring after the radiosurgery procedure. Clinical follow-up was obtained from a combination of clinic appointments and hospital admissions to the University of Virginia and correspondence with patients’ local referring physicians and outside hospitals. Patients were monitored with serial MRI studies every 6 months after radiosurgery for the first 2 years, followed by annual MRIs after the first 2 years. Neuroimaging in addition to routine follow-up imaging was obtained in patients with neurological decline. All follow-up imaging was reviewed by a neurosurgeon and a neuroradiologist at the University of Virginia, regardless of where it was obtained. AVM obliteration was defined on MRI by the absence of flow voids or on angiography by the lack of aberrant arteriovenous shunting. Angiography was typically reserved for AVMs that were determined to be completely obliterated on MRI. Hemorrhage was defined on CT or MRI with or without correlation to new or worsening neurological deficits. Radiation-induced changes (RIC) were defined on follow-up MRI by perinidal T2-weighted hyperintensities. RICs were classified as symptomatic if they correlated with clinical symptoms, most commonly headache, focal neurological deficit, or seizure. Statistical Analysis Statistical analysis was performed with the IBM SPSS
Radiosurgery seizure outcomes temporal AVMs
20 software program. Data are presented as mean and range for continuous variables and as frequency for categorical variables. For patients with seizures at presentation, the primary outcome variable was seizure control. For patients without seizures at presentation, the primary outcome variable was de novo seizures. The baseline AVM characteristics, radiosurgical treatment parameters, and postradiosurgery outcomes were compared between the temporal lobe AVM and non-temporal, cortical AVM cohorts using the unpaired Student t-test for continuous variables or Pearson chi-square test for categorical variables. Since the case and control cohorts had statistically similar AVM characteristics and radiosurgery treatment parameters, the 2 cohorts were merged for the univariate logistic regression analysis for predictors of seizure control. Covariates included in the logistic regression analysis were temporal lobe location, sex, age, prior AVM hemorrhage, prior embolization, prior resection, AVM volume, AVM location (non-eloquent vs eloquent), location of draining veins (exclusively superficial vs deep or both), number of draining veins (single vs multiple), prescription dose, Spetzler-Martin grade, RBAS, VRAS score, obliteration, radiological presence of RIC, postradiosurgery cyst formation, and postradiosurgery hemorrhage. Interaction and confounding was assessed through stratification and relevant expansion covariates. Factors predictive in univariate analysis (p < 0.20) were entered into a multivariate logistic regression analysis. A p value less than 0.05 was considered statistically significant. All statistical tests were 2-sided. For patients without seizures at presentation, the incidence of de novo seizures was too low to justify performing a logistic regression analysis. Nidus obliteration was reported as a cumulative percentage that combined obliteration confirmed by angiography and obliteration determined by MRI alone. KaplanMeier analysis was performed to determine the actuarial obliteration rates over time. The actuarial obliteration rates were calculated separately for patients with and without presenting seizures and were statistically compared with the log-rank test. The annual postradiosurgery hemorrhage risk was calculated by dividing the total number of latencyperiod hemorrhages by the total number of risk years. The number of risk years for each patient was defined as the time interval between radiosurgery and obliteration or last radiological follow-up for nonobliterated AVMs.
Results
Comparison of Baseline Characteristics of Case and Control Cohorts A total of 175 patients with temporal lobe AVMs met the inclusion criteria for analysis (Table 1). For the analysis of seizure control, the case cohort comprised 66 patients with temporal lobe AVMs who presented with seizures, and the control cohort comprised 122 patients with non-temporal, cortical AVMs who presented with seizures (Table 2). For the analysis of de novo seizures, the case cohort comprised 109 patients with temporal lobe AVMs who did not present with seizures, and the control cohort comprised 252 patients with non-temporal, cortical AVMs who did not present with seizures (Table 3). There were no
statistically significant differences in the AVM and treatment characteristics between the case and control cohorts for patients with or without seizure presentation. Effect of Temporal Lobe Location on Postradiosurgery Seizure Control Of the 66 temporal lobe AVM patients who presented with seizures, 5 (7.6%) had worsening of their seizures after radiosurgery. At last follow-up, seizure control was achieved in 41 patients (62.1%)—seizure freedom in 12 patients (18.2%) and decreased seizure frequency in 29 patients (43.9%). The remaining 20 patients had unchanged seizure outcomes following radiosurgery (30.3%). The rates of seizure freedom and control for patients with obliterated temporal lobe AVMs were 22.2% (8/36 patients) and 66.7% (24/36 patients), respectively. The rates of seizure freedom and control for patients with patent temporal lobe AVMs were 13.3% (4/30 patients) and 56.7% (17/30 patients), respectively (Fig. 1). Obliteration of the nidus did not significantly alter the rate of seizure freedom or control for patients with seizure presentation (p = 0.103 and p = 0.404, respectively). Of the 122 non-temporal, cortical AVM patients who presented with seizures, seizure control was achieved in 63 patients (51.6%), including seizure freedom in 18 patients (14.8%). The rates of seizure control and seizure freedom for patients with obliterated nidi were 52.9% (36/68 patients) and 16.2% (11/68 patients). Compared with patients with angioarchitecturally similar non-temporal cortical AVMs (Fig. 2A), patients with temporal lobe AVMs presenting with seizures did not have significantly different rates of seizure control (p = 0.168) or seizure freedom (p = 0.540). When only patients with completely obliterated nidi were considered, the differences between the 2 cohorts were also nonsignificant with respect to seizure control (p = 0.178) and seizure freedom (p = 0.448). Based on the univariate logistic regression analysis, preradiosurgery embolization (p = 0.023) and lower prescription dose (p = 0.027) were significant predictors of postradiosurgery seizure control (Table 4). In addition to these 2 covariates, temporal lobe location (p = 0.187), male sex (p = 0.120), preradiosurgery hemorrhage (p = 0.064), higher volume (p = 0.056), non-eloquent location (p = 0.087), and higher RBAS (p = 0.091) were selected for further analysis in the multivariate model. Multivariate logistic regression analysis did not identify any factors that were significantly associated with seizure control. Therefore, temporal lobe location was not significantly associated with seizure control. Additionally, obliteration was not associated with seizure control in the logistic regression analysis (p = 0.522). Effect of Temporal Lobe Location on De Novo Seizures After Radiosurgery Of the 109 temporal lobe AVM patients without seizures at presentation, 2 had de novo seizures after radiosurgery (1.8%) including 1.3% in obliterated (1/75 patients) and 2.9% in patent (1/34 patients) AVMs. AVM obliteration did not significantly alter the incidence of de novo seizures for patients without seizure presentation (p = 0.562). Of the 252 non-temporal, cortical AVM patients without seizure J Neurosurg April 17, 2015
3
D. Ding et al.
TABLE 1. Temporal lobe AVM cohort patient demographic characteristics, nidus characteristics, and treatment parameters Characteristic Sex Male Female Age (yrs) Mean Median Range Pediatric patients (age
Radiosurgery for temporal lobe arteriovenous malformations: effect of temporal location on seizure outcomes Dale Ding, MD,1 Mark Quigg, MD, MSc,2 Robert M. Starke, MD, MSc,1 Zhiyuan Xu, MD,1 Chun-Po Yen, MD,1 Colin J. Przybylowski, BS,1 Blair K. Dodson, BSN,1 and Jason P. Sheehan, MD, PhD1 Departments of 1Neurological Surgery and 2Neurology, University of Virginia, Charlottesville, Virginia
Object The temporal lobe is particularly susceptible to epileptogenesis. However, the routine use of anticonvulsant therapy is not implemented in temporal lobe AVM patients without seizures at presentation. The goals of this case-control study were to determine the radiosurgical outcomes for temporal lobe AVMs and to define the effect of temporal lobe location on postradiosurgery AVM seizure outcomes. Methods From a database of approximately 1400 patients, the authors generated a case cohort from patients with temporal lobe AVMs with at least 2 years follow-up or obliteration. A control cohort with similar baseline AVM characteristics was generated, blinded to outcome, from patients with non-temporal, cortical AVMs. They evaluated the rates and predictors of seizure freedom or decreased seizure frequency in patients with seizures or de novo seizures in those without seizures. Results A total of 175 temporal lobe AVMs were identified based on the inclusion criteria. Seizure was the presenting symptom in 38% of patients. The median AVM volume was 3.3 cm3, and the Spetzler-Martin grade was III or higher in 39% of cases. The median radiosurgical prescription dose was 22 Gy. At a median clinical follow-up of 73 months, the rates of seizure control and de novo seizures were 62% and 2%, respectively. Prior embolization (p = 0.023) and lower radiosurgical dose (p = 0.027) were significant predictors of seizure control. Neither temporal lobe location (p = 0.187) nor obliteration (p = 0.522) affected seizure outcomes. The cumulative obliteration rate was 63%, which was significantly higher in patients without seizures at presentation (p = 0.046). The rates of symptomatic and permanent radiationinduced changes were 3% and 1%, respectively. The annual risk of postradiosurgery hemorrhage was 1.3%. Conclusions Radiosurgery is an effective treatment for temporal lobe AVMs. Furthermore, radiosurgery is protective against seizure progression in patients with temporal lobe AVM–associated seizures. Temporal lobe location does not affect radiosurgery-induced seizure control. The low risk of new-onset seizures in patients with temporal or extratemporal AVMs does not seem to warrant prophylactic use of anticonvulsants. http://thejns.org/doi/abs/10.3171/2014.10.JNS141807
T
KEY WORDS epilepsy; Gamma Knife; intracranial arteriovenous malformation; stereotactic radiosurgery; seizures; temporal lobe; vascular malformations; vascular disorders
of temporal lobe arteriovenous malformations (AVMs) carries burdens beyond the primary consideration of prevention of hemorrhage. Prior studies of AVM-associated epilepsy have shown that higher rates of epilepsy are associated with temporal lobe involvement.20,23,24,48 Interventional therapy has the potential to control AVM-associated epilepsy.2,23 Since postreatment
treatment epilepsy status has a strong influence on quality of life and functional outcomes, treatment-induced seizure control is an important consideration.25 Thus, the success of an intervention is dictated by more than obliteration of the AVM nidus. Radiosurgery is a minimally invasive treatment alternative to microsurgery for AVMs and induces gradual obliteration of the nidus over
ABBREVIATIONS AED = antiepileptic drug; AVM = arteriovenous malformation; MTLE = mesial temporal lobe epilepsy; RBAS = radiosurgery-based AVM score; RIC = radiation-induced change; VRAS = Virginia Radiosurgery AVM Scale. submitted August 6, 2014. accepted October 20, 2014. include when citing Published online April 17, 2015; DOI: 10.3171/2014.10.JNS141807. DISCLOSURE The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper. ©AANS, 2015
J Neurosurg April 17, 2015
1
D. Ding et al.
a typical period of 2 to 3 years.37,47 While AVM radiosurgery outcomes have been previously reported by our institution and other neurosurgical centers, the relationship between temporal lobe location and AVM radiosurgery outcomes, specifically with regard to seizure control, has not been previously defined.9–11,27,33 Furthermore, despite the particular susceptibility of patients harboring temporal lobe lesions to seizures, antiepileptic drug (AED) therapy is not routinely administered to patients with temporal lobe AVMs who do not have seizures at presentation.14 Therefore, the goals of the present study are 1) to evaluate the radiosurgery outcomes, with respect to seizure control, obliteration, and complications, for temporal lobe AVMs and 2) to determine the effect of temporal lobe location on seizure control following AVM radiosurgery.
Methods
Patient Selection This is a single-center, retrospective, case-control cohort study evaluated from a prospectively collected database, approved by the University of Virginia institutional review board, of over 1400 AVM patients treated with Gamma Knife radiosurgery from 1989 to 2012. The inclusion criteria for the case cohort were 1) temporal lobe AVM location, 2) data regarding presence or absence of seizures at presentation, 3) data regarding posttreatment seizure outcomes available, 4) complete AVM obliteration following radiosurgery, and 5) minimum posttreatment follow-up duration of 2 years for AVMs without obliteration. Patients who had more than 1 radiosurgery treatment were not excluded, but the duration of follow-up was calculated from the last procedure. Radiosurgery Procedure The Gamma Knife (Elekta) radiosurgery procedure was performed in a standard fashion as previously described.47 Prior to 1991, MRI was not routinely used in addition to angiography for treatment planning. After 1991, a combination of MRI and angiography were routinely used to optimize the spatial accuracy of radiosurgical planning. From 1989 until June 1994, the Kula software was used for dose planning. From July 1994 onward, dose planning was performed with the GammaPlan software (Elekta). Data and Variables The following variables were obtained from the database and directed chart review: patient demographic characteristics (sex, age, and presenting symptom), baseline AVM characteristics (prior hemorrhage, prior embolization, prior resection, maximum diameter, volume, eloquent location, location and number of draining veins, and presence of associated aneurysms), and radiosurgery characteristics (prescription dose, maximum dose, isodose line, and number of isocenters). The Spetzler-Martin grade, modified radiosurgery-based AVM score (RBAS), and Virginia Radiosurgery AVM Scale (VRAS) were determined for each nidus.45,46,49 To determine the effect of temporal lobe location on AVM-associated seizure outcomes following radiosurgery, a control cohort of patients with non-temporal corti2
J Neurosurg April 17, 2015
cal AVMs was generated from the same institutional database. The inclusion criteria for the control cohort were the same as those for the case cohort with the exception of location (i.e., frontal, parietal, or occipital lobe location for the control cohort vs temporal lobe location for the case cohort). To prevent statistical bias due to differences in baseline AVM characteristics and treatment parameters, the control cohort was selected so that the frequencies of preradiosurgery hemorrhage, preradiosurgery embolization and resection, eloquent location, presence of deep venous drainage, number of draining veins, and presence of associated aneurysms and the means of the nidus volume and prescription dose did not differ significantly between the temporal lobe AVM and non-temporal, cortical AVM cohorts. The generation of the control cohort was performed blinded to outcome and was conducted separately for AVM patients presenting with seizures and AVM patients not presenting with seizures. Radiological and Clinical Follow-Up Information regarding seizure status was extracted from the chart and categorized as seizure free (with seizure freedom being patient-defined absence of seizures) or as not seizure free (patient-defined improved, unchanged, or worsened seizure rates). Seizure control was defined, in patients with seizures at presentation, as seizure freedom or decreased seizure frequency regardless of AED therapy. Seizure worsening was defined as increased frequency or intensity of seizures regardless of AED therapy. For the purposes of data analysis, postradiosurgery seizure outcomes were dichotomized into those patients with seizure control and those patients without seizure control (i.e., unchanged seizure status or worsening). De novo seizures were defined, in patients without seizure presentation, as new-onset seizures occurring after the radiosurgery procedure. Clinical follow-up was obtained from a combination of clinic appointments and hospital admissions to the University of Virginia and correspondence with patients’ local referring physicians and outside hospitals. Patients were monitored with serial MRI studies every 6 months after radiosurgery for the first 2 years, followed by annual MRIs after the first 2 years. Neuroimaging in addition to routine follow-up imaging was obtained in patients with neurological decline. All follow-up imaging was reviewed by a neurosurgeon and a neuroradiologist at the University of Virginia, regardless of where it was obtained. AVM obliteration was defined on MRI by the absence of flow voids or on angiography by the lack of aberrant arteriovenous shunting. Angiography was typically reserved for AVMs that were determined to be completely obliterated on MRI. Hemorrhage was defined on CT or MRI with or without correlation to new or worsening neurological deficits. Radiation-induced changes (RIC) were defined on follow-up MRI by perinidal T2-weighted hyperintensities. RICs were classified as symptomatic if they correlated with clinical symptoms, most commonly headache, focal neurological deficit, or seizure. Statistical Analysis Statistical analysis was performed with the IBM SPSS
Radiosurgery seizure outcomes temporal AVMs
20 software program. Data are presented as mean and range for continuous variables and as frequency for categorical variables. For patients with seizures at presentation, the primary outcome variable was seizure control. For patients without seizures at presentation, the primary outcome variable was de novo seizures. The baseline AVM characteristics, radiosurgical treatment parameters, and postradiosurgery outcomes were compared between the temporal lobe AVM and non-temporal, cortical AVM cohorts using the unpaired Student t-test for continuous variables or Pearson chi-square test for categorical variables. Since the case and control cohorts had statistically similar AVM characteristics and radiosurgery treatment parameters, the 2 cohorts were merged for the univariate logistic regression analysis for predictors of seizure control. Covariates included in the logistic regression analysis were temporal lobe location, sex, age, prior AVM hemorrhage, prior embolization, prior resection, AVM volume, AVM location (non-eloquent vs eloquent), location of draining veins (exclusively superficial vs deep or both), number of draining veins (single vs multiple), prescription dose, Spetzler-Martin grade, RBAS, VRAS score, obliteration, radiological presence of RIC, postradiosurgery cyst formation, and postradiosurgery hemorrhage. Interaction and confounding was assessed through stratification and relevant expansion covariates. Factors predictive in univariate analysis (p < 0.20) were entered into a multivariate logistic regression analysis. A p value less than 0.05 was considered statistically significant. All statistical tests were 2-sided. For patients without seizures at presentation, the incidence of de novo seizures was too low to justify performing a logistic regression analysis. Nidus obliteration was reported as a cumulative percentage that combined obliteration confirmed by angiography and obliteration determined by MRI alone. KaplanMeier analysis was performed to determine the actuarial obliteration rates over time. The actuarial obliteration rates were calculated separately for patients with and without presenting seizures and were statistically compared with the log-rank test. The annual postradiosurgery hemorrhage risk was calculated by dividing the total number of latencyperiod hemorrhages by the total number of risk years. The number of risk years for each patient was defined as the time interval between radiosurgery and obliteration or last radiological follow-up for nonobliterated AVMs.
Results
Comparison of Baseline Characteristics of Case and Control Cohorts A total of 175 patients with temporal lobe AVMs met the inclusion criteria for analysis (Table 1). For the analysis of seizure control, the case cohort comprised 66 patients with temporal lobe AVMs who presented with seizures, and the control cohort comprised 122 patients with non-temporal, cortical AVMs who presented with seizures (Table 2). For the analysis of de novo seizures, the case cohort comprised 109 patients with temporal lobe AVMs who did not present with seizures, and the control cohort comprised 252 patients with non-temporal, cortical AVMs who did not present with seizures (Table 3). There were no
statistically significant differences in the AVM and treatment characteristics between the case and control cohorts for patients with or without seizure presentation. Effect of Temporal Lobe Location on Postradiosurgery Seizure Control Of the 66 temporal lobe AVM patients who presented with seizures, 5 (7.6%) had worsening of their seizures after radiosurgery. At last follow-up, seizure control was achieved in 41 patients (62.1%)—seizure freedom in 12 patients (18.2%) and decreased seizure frequency in 29 patients (43.9%). The remaining 20 patients had unchanged seizure outcomes following radiosurgery (30.3%). The rates of seizure freedom and control for patients with obliterated temporal lobe AVMs were 22.2% (8/36 patients) and 66.7% (24/36 patients), respectively. The rates of seizure freedom and control for patients with patent temporal lobe AVMs were 13.3% (4/30 patients) and 56.7% (17/30 patients), respectively (Fig. 1). Obliteration of the nidus did not significantly alter the rate of seizure freedom or control for patients with seizure presentation (p = 0.103 and p = 0.404, respectively). Of the 122 non-temporal, cortical AVM patients who presented with seizures, seizure control was achieved in 63 patients (51.6%), including seizure freedom in 18 patients (14.8%). The rates of seizure control and seizure freedom for patients with obliterated nidi were 52.9% (36/68 patients) and 16.2% (11/68 patients). Compared with patients with angioarchitecturally similar non-temporal cortical AVMs (Fig. 2A), patients with temporal lobe AVMs presenting with seizures did not have significantly different rates of seizure control (p = 0.168) or seizure freedom (p = 0.540). When only patients with completely obliterated nidi were considered, the differences between the 2 cohorts were also nonsignificant with respect to seizure control (p = 0.178) and seizure freedom (p = 0.448). Based on the univariate logistic regression analysis, preradiosurgery embolization (p = 0.023) and lower prescription dose (p = 0.027) were significant predictors of postradiosurgery seizure control (Table 4). In addition to these 2 covariates, temporal lobe location (p = 0.187), male sex (p = 0.120), preradiosurgery hemorrhage (p = 0.064), higher volume (p = 0.056), non-eloquent location (p = 0.087), and higher RBAS (p = 0.091) were selected for further analysis in the multivariate model. Multivariate logistic regression analysis did not identify any factors that were significantly associated with seizure control. Therefore, temporal lobe location was not significantly associated with seizure control. Additionally, obliteration was not associated with seizure control in the logistic regression analysis (p = 0.522). Effect of Temporal Lobe Location on De Novo Seizures After Radiosurgery Of the 109 temporal lobe AVM patients without seizures at presentation, 2 had de novo seizures after radiosurgery (1.8%) including 1.3% in obliterated (1/75 patients) and 2.9% in patent (1/34 patients) AVMs. AVM obliteration did not significantly alter the incidence of de novo seizures for patients without seizure presentation (p = 0.562). Of the 252 non-temporal, cortical AVM patients without seizure J Neurosurg April 17, 2015
3
D. Ding et al.
TABLE 1. Temporal lobe AVM cohort patient demographic characteristics, nidus characteristics, and treatment parameters Characteristic Sex Male Female Age (yrs) Mean Median Range Pediatric patients (age
